Nelder–Mead Optimization of Dual-Wave EMAT for Defect Recognition With Grey Relational Analysis
Ting Zhang, Jieping Wu, Jiahao Zou, Wentao Peng, Xiaoqing Yang
Abstract
To enhance the capability of electromagnetic acoustic transducer (EMAT) in identifying defect types, this study employs the Nelder–Mead algorithm to optimize a dual-wave probe design for a rectangular sandwich structure. By exploiting the differential reflection intensities of shear and longitudinal waves on various defect types, the Grey relational analysis is utilized to analyze the echo amplitudes, facilitating the determination of defect types. Based on finite-element simulation, the design and optimization of the probe were accomplished. Simulation results demonstrate that the optimized probe’s energy conversion efficiency has increased by 137.5% compared to traditional designs. Comparative real measurements between the optimized and traditional probes confirm the high transduction efficiency of the designed structure. Twenty sets of defects, categorized into three types, were employed to evaluate the recognition capability, achieving a 100% identification rate. This offers an effective strategy to improve the recognition accuracy of defect types in electromagnetic ultrasonic detection.